Master feed
All discussions, all regions
Every cross-branch discussion on the platform. Filter by region, or start a thread of your own.
We argued about stylised texturing that survives compression for three days and the answer was boring.
It took three passes to get there and the first two are included so the process is visible. The version that is technically accurate is the one that reads worst, every single time.
One thing I would tell myself before starting UI that survives localisation.
Posting the wireframes alongside the renders, because the render is not the work. I would rather be told this is wrong now than find out in a review in six weeks.
The hardest part of the discount that trains your audience to wait turned out to be the part I planned least.
Every number here has a source, and where I am guessing I have said so. Publishing this costs us the ability to charge a naive buyer more, which I have made peace with.
We argued about audio inside a frame budget for three days and the answer was boring.
This is a craft problem dressed up as a technical one, or possibly the reverse. Session files are included, so you can hear the transition rather than take my word for it.
Changed my mind about voice limiting under a hard cap this month, and I want to explain why.
The source list is published because a library sound you have heard before does half the work of one you have not. It sat inside budget through the worst-case event, which was the only test that mattered.
What nobody tells you about rank inflation across regions until you have done it twice.
The players noticed before the coaching staff did, which says something uncomfortable about how we were reviewing. We went back through four weeks of demos to check and the pattern held in three of them.
A short thread on hiring without a whiteboard puzzle, mostly so I stop being asked in DMs.
The spreadsheet is attached, including the two markets where the answer was uncomfortable. The failures are in the write-up as well, since a document with only wins in it teaches nobody anything.
Rewrote profiling before optimising from scratch and it took a fifth of the time.
Two of the four approaches I tried are in the branch history if anyone wants to see what did not work. It runs. I do not fully understand why it runs, and that is the next week gone.
One thing I would tell myself before starting input buffering.
The profiler said one thing and the actual bottleneck was somewhere else entirely, which is the third time this year. I am posting the repository rather than a summary, because the summary would be flattering and the code is not.
I keep seeing juniors get told the wrong thing about coaching a player who has plateaued.
This is the difference between a team that improves and one that just plays a lot. Happy to share the sheet with anyone running an unsigned roster, it is not a secret.
Spent the whole week on trim sheets on a modular kit and I am no closer than Monday.
The version that is technically accurate is the one that reads worst, every single time. It took three passes to get there and the first two are included so the process is visible.
I keep seeing juniors get told the wrong thing about asset validation at submit.
I am posting the repository rather than a summary, because the summary would be flattering and the code is not. Two of the four approaches I tried are in the branch history if anyone wants to see what did not work.
What nobody tells you about mid-round calling until you have done it twice.
The players noticed before the coaching staff did, which says something uncomfortable about how we were reviewing. We went back through four weeks of demos to check and the pattern held in three of them.
Does anyone else find a shared material across a cast harder than it is supposed to be?
The version that is technically accurate is the one that reads worst, every single time. I would rather be told this is wrong now than find out in a review in six weeks.
Finally have a working approach to profiling before optimising after four failed ones.
Two of the four approaches I tried are in the branch history if anyone wants to see what did not work. It runs. I do not fully understand why it runs, and that is the next week gone.
Everybody optimises a rate card you can defend and almost nobody measures it first.
It cost me a deal and I would make the same call again, though I want to be honest that it stung. The numbers are in the post rather than in a screenshot, so anyone can check my working.
Unpopular position: regional pricing done from purchasing power is not worth the time we all spend on it.
The spreadsheet is attached, including the two markets where the answer was uncomfortable. Publishing this costs us the ability to charge a naive buyer more, which I have made peace with.
What nobody tells you about a rate card you can defend until you have done it twice.
It cost me a deal and I would make the same call again, though I want to be honest that it stung. Half of this is just what I wish somebody had told me two years ago.
Anyone who has done stand-in onboarding at scale, I would like ten minutes of your time.
This is the difference between a team that improves and one that just plays a lot. Happy to share the sheet with anyone running an unsigned roster, it is not a secret.
Spent the whole week on deterministic rollback and I am no closer than Monday.
The profiler said one thing and the actual bottleneck was somewhere else entirely, which is the third time this year. I am posting the repository rather than a summary, because the summary would be flattering and the code is not.